Chloroplast genes as genetic markers for inferring patterns of change, maternal ancestry and phylogenetic relationships among Eleusine species

Phylogenetic relationship between the nine species of Eleusine was investigated based on RFLP of the seven amplified chloroplast genes/intergenic spacers, trnK gene sequence and cpSSR markers. The maternal genome donor (E. indica, 2n=2x=18) of the allotetraploid (2n=4x=36, 2n=2x=38) Eleusine species, and the phylogenetic relationships between cultivated E. coracana (2n=4x=36) and wild species have been successfully resolved. The species-specific markers were also identified. The explicit identification of the maternal parent and that of the immediate wild progenitor of finger millet will be immensely useful for future genetic improvement and biotechnological program(s) of the crop species.


Introduction
The wild relatives of crop species often carry beneficial alleles that are effective against various biotic and abiotic stresses. They hold the key to successful crop improvement programmes through introgression of desired genes from wild to cultivated crop species (Dida and Devos 2006). In this context, the assessment of phylogenetic relationships at the inter-specific level and the identification of gene pools are considered important. Based on chloroplast deoxyribonucleic acid (cpDNA) diversity, we obtained information on the molecular phylogeny of finger millet (Eleusine coracana) vis-à-vis its wild relatives.
The genus Eleusine is a member of the tribe Eragrosteae, subfamily Chloridoideae and the family Poaceae. It is a small genus of nine species, which includes six diploid (2n ¼ 2x ¼ 16, 18, 20) and three polyploid (2n ¼ 4x ¼ 36, 38) species (Bisht and Mukai 2002;Liu et al. 2011). These species are widely distributed in the tropical and subtropical regions of Africa, Asia and South America (Phillips 1972). East Africa is considered to be the centre of diversity for the genus (Bisht and Mukai 2002).
Eleusine coracana (2n ¼ 4x ¼ 36), commonly known as finger millet or ragi, is the only economically important species of the genus. After sorghum and pearl millet, finger millet ranks third in cereal production in the semi-arid regions of the world (Bisht and Mukai 2002). It is widely cultivated in the arid and semi-arid regions of East Africa, India, Nepal and many other Asian countries for its grain and fodder value (Verma 2009). The grain is widely used for preparing bread, cakes, soup, puddings, porridge and fermented beverages (Hilu and deWet 1976;Chandrashekar 2010;Neves 2011). Finger millet is a rich source of essential amino acids and polyphenols, and it has comparatively higher levels of calcium and iron than other known cereals (Barbeau and Hilu 1993;Chandrashekar 2010). It has a number of medicinal properties as well, particularly in controlling blood sugar levels in diabetic patients (Duke and Wain 1981;Chandrashekar 2010;Pradhan et al. 2010).
The assessment of genomic relationships between E. coracana and its allied species has been a subject of comprehensive investigations at the morphological, cytogenetic, biochemical and DNA level. Chromosome research has demonstrated the significant role of polyploidy and aneuploidy in the evolution of the genus (Chennaveeraiah and Hiremath 1974;Hiremath and Chennaveeraiah 1982;Hiremath and Salimath 1991;Bisht and Mukai 2000, 2001a, b, 2002. Biochemical, nuclear and cpDNA markers have provided valuable insight into relationships, and on the origin of the crop species (Hilu et al. 1978;Hilu 1988Hilu , 1995Johnson 1992, 1997;Werth et al. 1993Werth et al. , 1994Salimath et al. 1995a;Neves et al. 2005;Dida et al. 2007Dida et al. , 2008Liu et al. 2011). The 2n number and the genomic formula proposed on the basis of earlier studies are given in Table 1.
In the present study, the cpDNA restriction site pattern variation of seven amplified chloroplast genes/intergenic spacers, the chloroplast trnK gene sequence and cp microsatellite polymorphism were investigated for the first time in all nine Eleusine species, with the objective of constructing the phylogeny of the genus Eleusine and identifying the maternal genome donors of the polyploid species including E. coracana. Both direct sequencing of the trnK gene and polymerase chain reaction -restriction fragment length polymorphism (PCR -RFLP) of cpDNA regions and chloroplast simple sequence repeats (cpSSRs) are considered to be very good markers for detecting cpDNA variation. The chloroplast trnK gene (which also contains the matK gene within the trnK gene intron) sequence has been effectively used for the construction of grass phylogenies . The matK gene sequence is one of the seven loci widely utilized for the DNA barcoding of plants (Babbar et al. 2012). Specific chloroplast genes and/or intergenic spacers can be amplified (Taberlet et al. 1991;Demesure et al. 1995;Tsumura et al. 1995Tsumura et al. , 1996Dhingra and Folta 2005;Heinze 2005). The amplicons can be directly sequenced or restriction endonuclease digested (PCR -RFLP or cleaved amplified polymorphic sequence), and the occurrence of microsatellites (cpSSRs) in the chloroplast genome has been widely utilized for species identification, reconstruction of phylogenetic relationships, taxonomic studies and the identification of maternal parents in polyploids (Tsumura et al. 1995(Tsumura et al. , 1996Weising and Gardner 1999;Ishii and McCouch 2000;Lakshmi et al. 2000;Parani et al. 2000Parani et al. , 2001Komatsu et al. 2001;Provan et al. 2001;Kishimoto et al. 2003;Nwakanma et al. 2003;Zhu et al. 2003;Asadi Abkenar et al. 2004Van Droogenbroeck et al. 2004;Ibrahim et al. 2007;Angioi et al. 2008;Sehgal et al. 2008;Jena et al. 2009;Liu et al. 2011;Poczai et al. 2011).

Plant materials
Seeds of the Eleusine species and one outgroup species (Dactyloctenium aegyptium) utilized in the present study were obtained from the United States Department of Agriculture (USDA) (Beltsville, MD, USA), the International Livestock Research Institute (ILRI) (Addis Ababa, Ethiopia), Dr Mathews M. Dida (Maseno University, Maseno, Kenya) and the National Bureau of Plant Genetic Resources (NBPGR) (New Delhi, India). The accession numbers and source of the seed material are given in Table 1. The seeds were grown under controlled conditions in the experimental field of the Department of Botany, University of Delhi.

DNA extraction
The total genomic DNA of the Eleusine species and D. aegyptium was isolated from fresh young leaves using the cetyl trimethyl ammonium bromide method as described by Murray and Thompson (1980) with some modifications. Instead of a CsCl -ethidium bromide ultracentrifugation step, the DNA was purified by phenol -chloroform extraction.  Bisht and Mukai (2002).

Species
Accession
For PCR-RFLP analysis, the presence (1) or absence (0) of a restriction fragment was recorded. Total and mean character differences between pairs of species were calculated using PAUP* 4.0 (Swofford 2002). Nei and Li's coefficient of genetic distance (Nei and Li 1979) was calculated between each pair of species after the optimality criterion was set to DISTANCE. Cluster analysis was carried out using the unweighted pair-group method using arithmetic averages (UPGMA) (Sneath and Sokal 1973) and neighbour-joining (NJ) (Saitou and Nei 1987) methods. Bootstrap values were calculated from 100 replicates using the BOOTSTRAP command in PAUP.

PCR amplification, cloning and sequencing of the trnK gene, and data analysis
The trnK gene was amplified from the nine Eleusine species and one outgroup as described above. The amplification products were separated on a 1 % agarose gel, excised from the gel and purified using a QIA quick gel extraction kit (Qiagen, Germany). The purified amplification products were cloned and sequenced with an ABI PRISM  Table 2. List of chloroplast genes/intergenic spacers amplified in the present study. a Sizes in source.

Genes/intergenic spacers
Primer pair Size in bp a Source Oryza sativa (Demesure et al. 1995) psaA Nicotiana tabacum (Shinozaki et al. 1986;Tsumura et al. 1996) 16S Nicotiana tabacum (Shinozaki et al. 1986;Tsumura et al. 1995Tsumura et al. , 1996 trnK Oryza sativa (Hiratsuka et al. 1989;Tsumura et al. 1995Tsumura et al. , 1996 psbD Nicotiana tabacum (Shinozaki et al. 1986;Tsumura et al. 1995Tsumura et al. , 1996 trnL-trnF Nicotiana tabacum (Taberlet et al. 1991) trnH-trnK Nicotiana tabacum (Demesure et al. 1995) 377 automated DNA sequencer (Applied Biosystems, Foster City, CA, USA). BLAST similarity searches were performed using the National Centre for Biotechnology Information (NCBI) BLASTN algorithm to confirm the identity of the trnK sequences. The nucleotide sequences have been submitted to GenBank with accession numbers KF357736 -KF357745. The nucleotide sequence of the partial trnK gene from Acrachne racemosa available in GenBank (accession number JN681616.1) was also utilized for phylogenetic analysis. The consensus sequence for the trnK gene from the nine Eleusine taxa and one outgroup (D. aegyptium) was generated using the software CLC Sequence viewer 6.1. The sequences were aligned using Clustal-X, version 1.8 ( Thompson et al. 1997). Phylogenetic analyses were carried out with the software MEGA 5 (Tamura et al. 2011). Pairwise sequence divergence rates between species were calculated using the maximum composite likelihood method. Phylogeny construction was carried out using NJ, minimum evolution (ME), maximum likelihood (ML) and maximum parsimony (MP) methods. Neighbour-joining and ME trees were obtained using the maximum composite likelihood criterion while ML and MP trees were constructed using the nearest-neighbourinterchange (NNI) and tree-bisection-reconnection algorithms, respectively. In all the analyses, all positions containing gaps and missing data were eliminated from the dataset (complete deletion option). Support values of the internal branches of NJ, ME, ML and MP trees were evaluated by bootstrap (500 replicates).

PCR amplification of chloroplast microsatellites, polyacrylamide gel electrophoresis and sequencing
A total of eight primer pairs (Table 3) were used for amplification of the chloroplast microsatellites. Amplification was carried out in 25 mL reaction mixtures containing 2.5 mL of 10× reaction buffer, 1.5 mM MgCl 2 , 200 mM dNTPs, 200 nM each primer, 0.5 U of Taq DNA polymerase (Bangalore Genei) and 25 ng of template DNA. DNA amplification was performed in a MyCycler TM (Bio-Rad) programmed to initial denaturation at 94 8C for 5 min followed by 35 cycles each of 1 min at 94 8C, 1 min at 55 8C (65 8C for ccmp2 and 53 8C for ccmp9) and 1 min at 72 8C, followed by a final extension cycle of 5 min at 72 8C.
An equal volume (10 mL) of formamide dye (98 % formamide, 10 mM EDTA, 0.026 g of bromophenol blue, 0.026 g of xylene cyanol) was added to each amplified product. The samples were heated for 5 min at 94 8C and immediately placed on ice. A total of 2.5 mL of each sample was loaded on a 6 % polyacrylamide gel (19 : 1 acrylamide : bisacrylamide, 7.5 M urea and 1× TBE buffer), and electrophoresis was conducted at 55 W and 55 8C for 2 h.
For silver staining, the gel was fixed in 10 % (v:v) acetic acid for 30 min. It was subsequently rinsed three times in de-ionized water (2 min per rinse). The gel was then kept for staining for 30 min in a 2 L solution containing 2 g of silver nitrate and 3 mL of 37 % formaldehyde (Promega, USA). The stained plate was rinsed with de-ionized water for 20 s and developed in a prechilled (10 8C) developer (2 L) solution containing 60 g of sodium carbonate, 3 mL of 37 % formaldehyde and 400 mL of sodium thiosulfate (10 mg mL 21 ). When bands became visible, the gel was immediately transferred to 10 % acetic acid solution to stop further reaction. The gel was finally rinsed with distilled water and air dried.
The PCR products of eight primer -template combinations were characterized by direct sequencing. A PCR was performed in 100 mL volumes as described above. An aliquot of the PCR product was checked by agarose gel electrophoresis, and the remainder was purified through a QIA quick PCR clean up kit (Qiagen). Nucleotide sequencing was performed using an ABI PRISM 377 automated DNA sequencer (Applied Biosystems). The nucleotide sequences have been submitted to GenBank with accession numbers KF357730 -KF357735.
Genetic distance (Nei and Li's) among the nine Eleusine species and one outgroup ranged from 0.0000 to 0.01857 (Table 5). Based on the restriction site data of the amplified gene/intergenic products, UPGMA and NJ dendrograms were generated. Both showed similar topologies. They resolved into two major clades. Clade I, supported by 100 % bootstrap support, contained E. kigeziensis, E. indica, E. tristachya, E. coracana and E. africana. Clade II was comprised of E. jaegeri, E. floccifolia, E. intermedia and E. multiflora. Eleusine floccifolia and E. intermedia were more closely related to each other (supported by a bootstrap value of 100 %) than either of the species was to E. jaegeri. Dactyloctenium aegyptium was the most diverged species among the analysed species (Fig. 2).

trnK sequence data
The trnK gene codes for tRNA-Lys UUU . In rice, the chloroplast trnK gene is 2576 bp long, of which the exon is only 72 bp long. The exon is divided into two parts by a long intron of 2500 bp. The 5 ′ exon consists of 37 bp and the 3 ′ exon consists of 35 bp. The matK gene, which codes for maturaseK, is 1536 bp in length and is located within the chloroplast trnK intron. In the present study, the trnK gene was sequenced from the nine Eleusine species and D. aegyptium. The aligned trnK gene sequence of Eleusine corresponded to position 45 at the 5 ′ end to position 2531 at the 3 ′ end of rice. The length of the trnK gene in the nine Eleusine species ranged between 2463 (E. jaegeri) and 2467 bp (E. tristachya and E. floccifolia).   Nipponbare (Ishii and McCouch 2000) and missing data comprised 2486 nucleotide positions, out of which 2254 were conserved, 214 were variable and 57 were parsimony informative sites. The trnK sequence data were correlated with the poly-morphic PCR-RFLP profiles of the trnK gene. Eight ampli-conenzyme combinations for the trnK gene showed   Combinations  I  VIII VIII  III  IV  I  VIII III  V  VI  VII VI  IV  VI  II  VIII VIII VIII  VIII VIII VIII VIII VIII polymorphic profiles which revealed that polymorphism in all the profiles was the result of site mutations. The entire trnK gene sequence from the taxa analysed showed 214 single-nucleotide polymorphisms (SNPs). A total of 10 SNPs in the restriction sites were responsible for the eight polymorphic profiles. The polymorphic profiles produced were due to gain or loss of the restriction sites caused by SNPs. The consensus trnK gene sequence was generated from the taxa analysed and the locations of the polymorphic sites were marked on it (Fig. 3).
Pairwise sequence divergence ranged from 0.003 to 0.058 (Table 6). Dendrograms were generated using four different methods (NJ, ML, ME and MP). The phylogeny reconstruction through the NJ method resulted in an optimal tree with a sum of branch length of 0.095. In the NJ bootstrap consensus tree, all the Eleusine species were grouped into two distinct clades with 100 % bootstrap support. Clade I consisted of E. coracana, E. africana, E. indica, E. kigeziensis and E. tristachya. Eleusine coracana and E. africana grouped together in one subclade with a bootstrap support of 95 %, and E. indica and E. kigeziensis grouped together in another subclade with a bootstrap support of 54 % within Clade I. Eleusine tristachya did not group within any of the two subclades and thus was the most diverged among the five species of Clade I with a bootstrap support of 98 %. Clade II consisted of E. intermedia, E. floccifolia, E. jaegeri and E. multiflora. Eleusine intermedia and E. floccifolia were more closely related to each other than to E. jaegeri with a bootstrap support of 54 %. Eleusine multiflora represented the most diverged species within Clade II. Acrachne racemosa and D. aegyptium were the most diverged species among all the species (Fig. 4A).
The robustness of the present data lies in the fact that the dendrograms based on ME, ML and MP analysis showed almost the same topology as that of the NJ tree (Fig. 4). Maximum parsimony analysis resulted in most parsimonious trees (Fig. 4C) with a tree length of 232. The bootstrap consensus tree had a consistency index of 0.943, a retention index of 0.859 and a rescaled consistency index of 0.811.

Chloroplast microsatellite polymorphism and sequence data
Eight consensus primer pairs were used to amplify cpDNA microsatellites from the nine species (Table 3). The primer pairs were able to amplify a product in all 10 species (Fig. 5). A major band was produced for each primertemplate combination. The typical stuttering phenomenon which usually occurs upon amplification of mononucleotide-and dinucleotide-type microsatellites was observed. Size variation for the amplified products was only observed for two cpSSR loci. Two size variants were detected for ccmp2 and ccmp5. For ccmp2, all but one (E. tristachya) species showed a 197-bp allele. In E. tristachya, it was 200 bp. In the case of ccmp5, E. jaegeri showed a band of 146 bp, while each of the remaining nine species had a 145-bp allele (Fig. 5).
To test the presence of a repeat motif in Eleusine, nucleotide sequencing was carried out for amplified products from E. coracana with all eight cpSSR primer sets. RCt5 and RCt8 could not be successfully sequenced. ccmp2, ccmp5, ccmp6, RCt3, RCt4 and RCt7 showed sequence sizes of 197, 145, 96, 113, 131 and 126 bp, respectively (Tables 3 and 7). Repeat motifs were identified in all six amplification products at the expected positions but the degree of conservation was variable (Table 7). Among the repeat motifs, RCt4 and RCt7 were highly conserved, ccmp5 and ccmp6 were moderately conserved, and ccmp2 and RCt3 were least conserved. In the case  of ccmp2 and RCt3, the number of mononucleotide repeats was reduced.

Discussion
The restriction site variation identified in the amplified chloroplast gene/intergenic spacers has provided new insights into the origin and evolution of the three polyploid species, and the genetic relationships between the cultivated and wild Eleusine species. Species-specific markers were also identified. None of the amplicon -enzyme combinations could discriminate between E. coracana (AABB, 2n ) and E. kigeziensis (AADD, 2n ¼ 4x ¼ 38) on the one hand, and between E. floccifolia (BB, 2n ¼ 2x ¼ 18) and E. intermedia (AB, 2n ¼ 2x ¼ 18) on the other hand. Although the maximum number of polymorphic and phylogenetically informative markers was obtained in eight trnK amplicon -enzyme combinations, the amount of variation that can be tapped with the PCR -RFLP method was lower than the variation tapped with sequencing of the trnK gene. A total of 214 SNPs were found in the trnK gene sequence, of which only 10 were causative to 8 polymorphic profiles. This clearly demonstrates that sequencing of the trnK gene is more informative than the PCR -RFLP method. Moreover, the nucleotide sequence of the chloroplast trnK gene fine-tuned the species AoB PLANTS www.aobplants.oxfordjournals.org  relationships. The nine species of Eleusine were grouped into two clades, in both PCR -RFLP and trnK gene sequence data analyses. The trnK sequence data clearly showed that the three tetraploids were closer to E. indica than to E. tristachya, as concluded from PCR -RFLP data. The present observations based on restriction site variation and the trnK gene sequence of cpDNA are congruent with the conclusions reached by various authors based on the multiple marker nuclear DNA assay, chloroplast markers and chromosome research. The strong affinities between E. coracana, E. africana and E. indica have been highlighted in earlier studies Johnson 1992, 1997;Hilu 1995;Mukai 2000, 2001a, b;Neves et al. 2005;Dida et al. 2007Dida et al. , 2008Liu et al. 2011) as well. A number of studies based on chromosome research (Hiremath and Chennaveeraiah 1982;Hiremath and Salimath 1991), 2C DNA content (Hiremath and Salimath 1991), crossability data (Salimath et al. 1995b) and ribosomal DNA polymorphism (Hilu and Johnson 1992;Werth et al. 1994;Bisht and Mukai 2000), ITS sequence data (Neves et al. 2005) and cpDNA (Hilu and Johnson 1997;Neves et al. 2005;Liu et al. 2011) also support the close affinity between E. indica and E. tristachya. The allotetraploid E. kigeziensis exhibits the same maternal lineage as that of the two other allotetraploid species, E. coracana and E. africana. The five species share some common morphological features, such as 1 -3 nerved lower glumes with a winged keel, 3-7 nerved upper glumes with a more or less winged keel, the presence of 1-3 subsidiary nerves adjacent to the central nerve of the lemma and the generally winged keel of the palea (Phillips 1972).
The present data unambiguously support the view of Chennaveeraiah and Hiremath (1974), Hilu (1988), Hiremath and Salimath (1992), Hilu and Johnson (1997), Mukai (2000, 2001a, b) and Neves et al. (2005) that E. africana is the wild progenitor of the cultivated species, E. coracana, and that E. indica is one of the diploid progenitors of the two polyploid species (Liu et al. 2011). The present results further indicate that E. indica with the AA genome might be the maternal parent for all three tetraploid species, viz. E. coracana, E. africana and E. kigeziensis. Morphologically, E. kigeziensis resembles E. indica. It is considered to be a hybrid between E. indica and one of the perennial species (Phillips 1972). Salimath (1990) and Bisht and Mukai (2002) have also proposed E. indica as one of the genome donors of E. kigeziensis.
The close affinity in the cpDNA PCR -RFLP profiles and trnK sequences between E. floccifolia and E. intermedia is also supported by the occurrence of the same chromosome number (2n ¼ 18), 2C DNA values (Hiremath and Chennaveeraiah 1982;Hiremath and Salimath 1991) and ITS sequence data (Neves et al. 2005). Close association between E. jaegeri, E. floccifolia and E. intermedia suggested in earlier studies (Hiremath and Salimath 1991;Hilu and Johnson 1992;Liu et al. 2011) is further supported by our data. Eleusine jaegeri in all combinations grouped together with E. floccifolia and E. intermedia except in the psaA -MspI combination, where it shows a unique profile vis-à -vis the remaining species. In the case of trnK sequence analysis also these three species are found to be closely associated. Eleusine multiflora also behaves in the same fashion. The species exhibits unique profiles vis-à -vis the remaining species in trnK -DraI, trnK -MspI, trnK -SpeI, psaA-TaqI and psbD -HaeIII combinations, with the result that it represents the most diverged taxon among all the Eleusine species. These amplicon -enzyme  1986)], distinct chemical composition (Hilu et al. 1978), molecular data (Hilu and Johnson 1992;Hilu 1995), chromosome number (2n ¼ 16), 2C DNA value and other cytogenetic features (Mysore and Baird 1997;Bisht and Mukai 2000) including the present results. Dactyloctenium aegyptium in restriction site analysis, and A. racemosa and D. aegyptium in trnK sequence analysis indicate that they are the most diverged species vis-à -vis Eleusine species analysed and thus represented as outgroups.
Chloroplast microsatellites are known to have potential for phylogenetic studies (Powell et al. 1995a(Powell et al. , b, 1996Provan et al. 1997Provan et al. , 2001Ishii and McCouch 2000;Angioi et al. 2008). The present cpSSR data on Eleusine species, however, were not helpful as very little polymorphism was obtained for various amplified microsatellites. Only two cpSSR loci (ccmp2 and ccmp5) were found to be polymorphic, displaying a total of four alleles. In the case of ccmp2, the variant allele was found in E. tristachya, while for ccmp5 the variant allele was found in E. jaegeri. Overall, the sequences seem to be highly conserved. This is in strong contrast to the observations made in the other genera such as Glycine, Oryza and Hordeum, where different species within the genus and even within different subspecies displayed characteristic haplotypes (Powell et al. 1995b(Powell et al. , 1996Provan et al. 1997Provan et al. , 1999. The reason for not obtaining polymorphism in the present study could be attributed to the fact that it is very difficult to design universal primers for microsatellites that show widespread polymorphism at the inter-specific level and at the same time are able to amplify across a broad range of plant genera. Although there are primers available that amplify across wide-ranging taxa, these rarely show widespread inter-specific polymorphism. This could be due to the contrasting requirements in a short stretch of DNA of extreme stability and sequence conservation for priming sites and consistently high levels of polymorphism in the intervening region (Provan et al. 2001). Second, size homoplasy may not necessarily mean identical intervening microsatellite regions; it may also be due to variation in flanking regions. It is also possible that back mutations have taken place, which might result in Figure 5. Silver-stained polyacrylamide gel electrophoresis patterns generated by ccmp5 (A) and ccmp6 (B) in E. coracana (1), E. africana (2), E. tristachya (3), E. indica (4), E. floccifolia (5), E. intermedia (6), E. kigeziensis (7), E. multiflora (8), E. jaegeri (9) and D. aegyptium (10). The size of the fragments in base pairs is indicated on the left. Table 7. Nucleotide sequences of cpDNA of E. coracana amplified using six cpSSR primer pairs. The location of repeats is underlined. TTTTTTCTGTCATGTTT-3 ′ identical size but not identical sequence (Bryan et al. 1999).

Conclusions
Our results based on RFLP of the seven amplified chloroplast genes/intergenic spacers, and the trnK gene sequence in the nine diploid and allotetraploid Eleusine species and two outgroup species resulted in well-resolved phylogenetic trees. The maternal genome donor (E. indica, 2n ¼ 2x ¼ 18) of the allotetraploid (2n ¼ 4x ¼ 36, 2n ¼ 2x ¼ 38) Eleusine species, and the phylogenetic relationships between cultivated E. coracana (2n ¼ 4x ¼ 36) and wild species could be successfully resolved. The speciesspecific markers were also identified. The two diploid species E. indica and E. tristachya could not be resolved separately by PCR -RFLP of seven chloroplast genic/ intergenic spacers, as not a single site change could be scored. However, the trnK gene sequence clearly demonstrated that E. indica is more closely related to all three allotetraploids as compared with E. tristachya. Therefore, E. indica is most likely the maternal parent to all three allotetraploids. Eleusine multiflora (2n ¼ 2x ¼ 16) was found to be the most diverged among all the species. The explicit identification of the maternal parent and that of the immediate wild progenitor of finger millet will be immensely useful for future genetic improvement and biotechnological programmes of the crop species.

Sources of Funding
This work was supported by the Council of Scientific and Industrial Research (CSIR), Government of India and the National Academy of Sciences, India (NASI).

R.
A. was involved in planning and performing all the experiments, data analyses and manuscript writing. N.A. was involved in the data analysis. R.T. was involved in relevant research discussions. S.N.R. was involved in the planning and supervision of all the experimental work and in writing the manuscript. All authors have seen and agreed to the submitted manuscript.

Conflicts of Interest Statement
None declared.